Welcome to SimpleSR’s documentation!

SimpleSR is a small and simple library for Deep Learning based Single Image Super-Resolution. Super Resolution is the field of study for increasing the resolution of images, while ideally preserving (or generating) finer texture details.

SimpleSR was developed as part of my Undergrad Thesis with the main focus to learn more about Super-Resolution and Deep Learning in general. Additionally I hope this library can serve as a starting point for others to experiment with Super-Resolution.

Note

The project is still under active development, things might not work or possibly break in the future. Feel free to open an issue if you encounter a problem.

Source code is available at github.

Getting started:

• Installation
  • Running natively
  • Running inside docker container
• Usage
  • General Principles
  • Training Models
  • Evaluating Models
• Configuration Examples

Results Gallery:

• Results Gallery
  • Examples from thesis

Module Reference:

• Data Pipeline
• Models
  • SRModel
  • Generator
  • Discriminator
• Training
• Evaluation
• Utils
  • Config Util
  • Image Util Module
  • Model Utils